[vtk-developers] xdmf questions (UNCLASSIFIED)

[John Biddiscombe]

Jerry,

Thanks for this information. I had a peek at the xdmf list archives as 
suggested by Berk and have digested what you've written below.

#define XDMF_TIME_SINGLE    0x00
ok - this seems straightforward

#define XDMF_TIME_UNIFORM   0x01
ok, a range of values specified as a vector

#define XDMF_TIME_HYPERSLAB 0x02
Not sure what this will be. Why not also XDMF_TIME_RANGE to allow a dataset to be valid for all times between t1,t2. Is this functionality going to be provided by TIME_FUNCTION or perhaps by Hyperslab.

#define XDMF_TIME_FUNCTION  0x03
ditto.

If several meshes exist inside a dataset and some have time values and 
others do noit, will the default behaviour be to assume those with no 
time attribute exist for all time values requested?

I need to start using this stuff as of yesterday-ish. Do I use the 
/cvsroot/Xdmf reporitory at kitware to get it, or is it mirrored in 
pv/vtk-utilities etc?

I am more than willing to help out, as I need to transfer a ton of 
binary fortran data into something more sensible and I'd like it to be 
hdf5 with xdmf wrappers. I will be working on the vtkXdmfReader 
andWriter classes and will add time reading support if nobody else is on 
it already.

Please feel free to transfer this discussion to an alternative list...

Thanks

JB




> Classification:  UNCLASSIFIED 
> Caveats: NONE
>
> John,
>
> I am currently adding a time facility to Xdmf. 
> I'll update the wiki @ xdmf.org when I am finished
> but basically time will be an additional element
> which is a child of Grid. In simplest form :
>
> <Grid GridType="Uniform" ...
>      <Time Value="0.1" />
>
> There is an XdmfTime object. time->GetTimeType() will return
> XDMF_TIME_SINGLE in the above example. Others are :
>
> #define XDMF_TIME_SINGLE    0x00
> #define XDMF_TIME_UNIFORM   0x01
> #define XDMF_TIME_HYPERSLAB 0x02
> #define XDMF_TIME_FUNCTION  0x03
>
> More complex :
> <Grid  GridType="Collection" ....
>              <Time TimeType="Uniform">
>                  <DataItem Format="XML" Dimensions="5" DataType="Float">
>                      1.1 2.2 3.3 4.4 5.5
>                  </DataItem>
>              </Time>
> Both of these are working, as is :
> 	<Time Function="???" />
> the time object doesn't do anything with it yet other than return a
> string.
>
> Here's some python from the test :
>
> print 'Time Type = ', tt, ' : ', t.GetTimeTypeAsString() if(tt ==
> XDMF_TIME_SINGLE) :
>      print 'Time = ', t.GetValue()
>      t.SetValue(99.0)
> elif(tt == XDMF_TIME_FUNCTION) :
>      print 'Function = ', t.GetFunction()
>      t.SetFunction('Hello')
>      print 'Function = ', t.GetFunction()
> elif(tt == XDMF_TIME_UNIFORM) :
>      a = t.GetArray()
>      print 'Values = ', a.GetValues()
>      a.SetValueFromFloat64(3, 99.0)
>
> t.Build()
> print t.Serialize()
>
>  
>
> -----Original Message-----
> From: vtk-developers-bounces+clarke=arl.army.mil at vtk.org
> [mailto:vtk-developers-bounces+clarke=arl.army.mil at vtk.org] On Behalf Of
> John Biddiscombe
> Sent: Wednesday, January 23, 2008 11:17 AM
> To: paraview at paraview.org; vtk-developers
> Subject: [vtk-developers] xdmf questions
>
> I am aware that others (who?) are working with xdmf and I have two
> specific questions I'd like to ask before diving in to make changes to
> read/writers
>
> 1) Has anyone given any consideration to time representation withing
> xdmf. Specifically, multiple time steps and the ongoing time support in
> vtk/paraview I need to store multiple time steps preferably in the same
> file, but if one step per hdf5 file could be referenced from the xdmf
> xml that's be ok I suppose. I don't see ant references to TIME_etc vars
> in the code. 
> If this is in development please let me know so that I can avoid
> duplication.
>
> 2) Data spaces and parallel write. I'd like to modify the xdmf writer to
> allow me to write pieces of volume data one by one  and specify the
> extents so that 8/16/64 blocks etc could be written into a single file.
> Has anyone started work on this? I wuld again like to avoid duplication.
>
> Many thanks
>
> JB
>
>   


-- 
John Biddiscombe,                            email:biddisco @ cscs.ch
http://www.cscs.ch/about/BJohn.php
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Cantonale, 6928 Manno, Switzerland      | Fax:  +41 (91) 610.82.82